libvpx: Pull from upstream

source/libvpx/vpx_dsp/x86/inv_wht_sse2.asm

 ;
 ;  Copyright (c) 2015 The WebM project authors. All Rights Reserved.
 ;
 ;  Use of this source code is governed by a BSD-style license
 ;  that can be found in the LICENSE file in the root of the source
 ;  tree. An additional intellectual property rights grant can be found
 ;  in the file PATENTS.  All contributing project authors may
 ;  be found in the AUTHORS file in the root of the source tree.
 ;
+
 %include "third_party/x86inc/x86inc.asm"
 
 SECTION .text
 
 %macro REORDER_INPUTS 0
   ; a c d b  to  a b c d
   SWAP 1, 3, 2
 %endmacro
 
 %macro TRANSFORM_COLS 0
@@ skipping 40 matching lines @@
   punpcklwd       m0, m1
   punpckhwd       m3, m1
   mova            m2, m0
   punpcklwd       m0, m3
   punpckhwd       m2, m3
   pshufd          m1, m0, 0x0e
   pshufd          m3, m2, 0x0e
 %endmacro
 
 %macro ADD_STORE_4P_2X 5  ; src1, src2, tmp1, tmp2, zero
-  movq            m%3,       [outputq]
-  movq            m%4,       [outputq + strideq]
+  movd            m%3,       [outputq]
+  movd            m%4,       [outputq + strideq]
   punpcklbw       m%3,       m%5
   punpcklbw       m%4,       m%5
   paddw           m%1,       m%3
   paddw           m%2,       m%4
   packuswb        m%1,       m%5
   packuswb        m%2,       m%5
   movd            [outputq], m%1
   movd            [outputq + strideq], m%2
 %endmacro
 
@@ skipping 11 matching lines @@
   TRANSPOSE_4X4
   REORDER_INPUTS
   TRANSFORM_COLS
 
   pxor            m4, m4
   ADD_STORE_4P_2X  0, 1, 5, 6, 4
   lea             outputq, [outputq + 2 * strideq]
   ADD_STORE_4P_2X  2, 3, 5, 6, 4
 
   RET